| | | 20,000+ Fresh Resumes Monthly | |
|
|
| |
Click here or scroll down to respond to this candidateCandidate's Name
Brookline, MA PHONE NUMBER AVAILABLE PHONE NUMBER AVAILABLE EMAIL AVAILABLEPROFESSIONAL SUMMARYPrincipal data and machine learning scientist responsible for pioneering an AI/ML driven approach to telecommunications and IoT operations support systems. Assembled mission critical, cloud native, AI/ML platforms and solutions. Pioneered the application of analytical, computational and AI/ML methods to problems lying at the intersection of telecommunications, high performance computing and high-performance networking domains. Thrives in cross-functional collaboration and working on multi-disciplinary problems. Advised the business and management on strategic go-to-market initiatives resulting in increased revenues and operating margins. Translated customer business challenges into actionable insights by applying the complete data science workflow, which resulted in increased revenues and operating expense savings. Holds several publications in international journals, multiple granted and pending patents, and presentations in international conferences. Looking for opportunities that will leverage both technical and business acumen to deliver creative products and solutions enabling organizations to benefit from increased revenues and operating margins. Core competencies include: Artificial Intelligence / Machine Learning Deep Learning (CNN/FCN/RNN/LSTM/Transformer) Generative AI / LLM / RAG / Fine-Tuning / LoRA / QLoRA Data Science Optimization NLP/NLU (BERT) Probabilistic Graph Networks Simulation Quantitative Research PyTorch Tensorflow Huggingface Langchain C/C++/Java/Scala/Python/SQL Data Engineering and Analysis Hadoop / Map-Reduce / Hive Apache Spark MLLib Sci-Kit Learn gensim FAISS EXPERIENCEHewlett Packard Enterprise November 2007 - Present Distinguished Technologist June 2021 - Present Served as a trusted advisor to service providers across the world (AT&T, Verizon, Reliance Jio, Vodafone Idea, Optus, Vodafone NZ) strategizing and delivering creative business and technical solutions driven by cutting edge AI/ML. Leveraged Generative AI (LLMs and SLMs) to develop intelligent agents assisting network operations in detecting and diagnosing network problems, displaying actionable insight with possible remediation steps. o Conversational AI applications for product documentation and customer incident cases using RAG to generate accurate LLM responses.o Fine-tuned LLM application to replace legacy rule based expert system for root cause inference in telecommunication network problem detection use cases. Products and solutions enabled customers to reduce service outages and equipment failures, optimize energy utilization, spectrum utilization, and operations processes, by adopting a proactive approach. Annualized software and services revenues exceeding 50 million dollars generated. Lead Technologist November 2007 - June 2021Worked with large scale systems that are critical to our customers' business. Applied AI/ML, optimization, high performance computing, and analytical skills to customer problems across multiple industries. Built long term technology and system roadmaps, which align with product roadmaps and address mission critical 24x7 operations, availability, scalability, and capacity. Skilled at using advanced statistical analysis to analyze massive data sets, apply machine learning / deep learning, to assemble models supporting anomaly detection, forecasting, classification of incidents and recommend remediation steps. Proposed and implemented solutions that have helped customers reduce operating costs by 40%, improved time- to-market and time-to-revenue for new services by 35%, reduced capital expenses in network equipment by 20%, allowed high value premium services to be delivered by 40%. One large US customer had operations expense savings exceeding seven billion dollars over a period of ten years. ADDITIONAL EXPERIENCEHewlett Packard Enterprise - Senior System Architect; System Architect; Engineer / Technology Consultant Supercomputer and Education Research Center, Indian Institute of Science and JNCASR - Research Scientist / Research AssociateEDUCATIONMaster of Business Administration MBA, MIT Sloan School of Management Ph.D., Mathematical Physics and Computational Science, Indian Institute of Science (IISc) PUBLICATIONS PATENTS 4 patents granted; 6 other patents filed (patents pending). 6 HPE Tech Con posters, 3 invited HPE Tech Con presentations. 10 publications in international refereed journals. Multiple conference presentations.WEBSITES FOR REFERENCE https://LINKEDIN LINK AVAILABLE https://scholar.google.com/citations?user=ozw3GroAAAAJ&hl=en https://www.credly.com/users/Candidate's Name
RECENT AI/ML EXPERIENCE1. Development of a predictive maintenance algorithm to detect deteriorating performance in Y-cables of Layer-2 switches, well in advance of failure.2. Design and implementation of an adaptive and self-healing system for optimization of tunnel- based routing, to re-locate IP backbone traffic to new private peering points. 3. Discovery of network outage signatures in call disconnect codes contained in CDRs. Operationalization of discovered signatures for detection in real-time using supervised learning algorithm. 4. Development of a deep learning time series classifier to forecast / predict radio link failures from historical region-wise, radio link performance and weather data. 5. System to score actionable tasks based on: Data and VoLTE Retainability, Data and VoLTE Block Percentage, Downlink Throughput, Packet Discard Rate, CQI measures. Scores are used to tag sites/tickets for business and technical impact and prioritized accordingly. Sites tagged with these scores allow operations teams to prioritize and organize their efforts.6. Development of a time series deep learning classifier to identify if anomalous signals in the harmonic sectors of radio signals received at a cell site are due to FM/Cable TV harmonic interference or due to weather conditions. 7. Using graph search algorithms, working with OTDRs, to diagnose and isolate problems in metropolitan area outside plant optical fiber installations.8. Applied Graph Convolutional Neural Networks, coupled with an LSTM or a Transformer, to predict network traffic at cell sites, allowing radio access network elements / functions to be scaled-in or scaled-down, or their frequency bands to be re-configured, or CPUs supporting network function workloads to be transitioned from P-states or Turbo Boost states (frequency scaled states) to C/S-states (sleep states), to reduce energy consumption.9. Self-learning algorithmic approach to learn network observability data patterns, using a combination of LLM based document clustering, unsupervised frequent pattern mining, and probabilistic graphical networks to derive root cause from a host of symptoms derived from multiple data sources. 10. Autonomous multivariate anomaly detection in network observability data using a combination of multiple unsupervised clustering algorithms, ensembling, and supervised learning using Random Forest, Gradient XG- Boost, MLP and CNN.11. Anomaly detection in univariate time series using multiple approaches: statistical methods applied to time series, statistical learning approaches, auto-encoder / decoder neural networks, LSTM. 12. 5G IoT device energy consumption categorization, using unsupervised learning, and optimization, using a reinforcement learning approach.13. Leveraging Generative AI (LLMs and SLMs) to develop intelligent agents that can assist network operations in detecting and diagnosing network problems significantly faster, coming up with actionable insight and suggesting possible remediation steps.14. Used NLP/NLU (BERT) and vector indexed search algorithms to assemble semantic search engine for network and system generated logs.15. Conversational AI applications for product documentation and customer incident cases using RAG to generate accurate LLM responses.16. Fine-tuned LLM applications to replace legacy rule based expert system for root cause inference in telecommunication network problem detection use cases. In some cases used LoRA and QLoRA. |
